We also conducted genome-wide scans for associations between gene expression traits and unlinked SNPs. Such trans-eQTLs may represent regulatory interactions between transcription factors, signaling molecules, or chromatin regulators and their target genes. After adjusting for demographic variables as above, we found 353 gene expression traits with significant (BF>5) trans-linked genetic effects. The replication behavior of trans-eQTLs was markedly different from cis-eQTLs (compare Figure 3B, 3C with Figure 3E, 3F). First, the distribution of t-test p-values derived from the UW replication set, for each best associated gene-SNP pair identified in the UC set, was effectively uniform (Figure 3E). Second, in contrast to cis-effects, which rapidly approach an asymptotic replication rate at BF 5, trans-eQTLs almost completely failed to replicate (6.14%; Figure 3F) at a BF threshold of 5. At greater significance thresholds, trans effects did replicate more frequently (e.g., at BF> = 9.5, 50.0% replicate), however, these rates never approached those observed for cis-eQTLs. It is plausible that surrogate variable correction may mask true ‘master’ regulator effects, but as for cis-effects we identified more trans-eQTLs with surrogate variable correction than without
